Deneb Samples are almost out

Printable View

Show 100 post(s) from this thread on one page

11-04-2008, 04:52 PM
stangracin2

Quote:

Originally Posted by Hornet331

actually you just did. :ROTF:

i was making sure someone told him
11-04-2008, 05:03 PM
Clairvoyant129

Quote:

Originally Posted by Bellisimo

imo, this is because the IMC was on the NB, L2 caches helps to hide the memory latency (L3 in nehalem and K10/K10.5)

but i really like to think about deneb as a RV770, which means a RV670 done right
i dont expect it to beat nehalem, but if it is a K10 done right, it may end op close in some scenarios

at least deneb will offer full control over the cpu, unlike nehalem where only the extreme part is usefull for overclocking and high mem speeds

:rolleyes:

I see you're spreading more of your fud. Francois already explained that there is no overclocking protection and what everyone is calling "overclocking protection" can be turned off from your your BIOS.

Anyway, if Deneb is competitively priced and offers 5%-15% performance improvement over Agena (clock for clock), I think it will be a good alternative to Yorkfield and Nehalem for the AMD fans.
11-04-2008, 05:14 PM
Glow9

Quote:

Originally Posted by Clairvoyant129

:rolleyes:

I see you're spreading more of your fud. Francois already explained that there is no overclocking protection and what everyone is calling "overclocking protection" can be turned off from your your BIOS.

Anyway, if Deneb is competitively priced and offers 5%-15% performance improvement over Agena (clock for clock), I think it will be a good alternative to Yorkfield and Nehalem for the AMD fans.

I think so too. Especially everyone who has a supporting motherboard already :up:
11-04-2008, 05:16 PM
Clairvoyant129

Quote:

Originally Posted by Glow9

I think so too. Especially everyone who has a supporting motherboard already :up:

Yup :up:
11-04-2008, 06:00 PM
AbelJemka

Quote:

Originally Posted by Shintai

You asked if it was available. It was.

It was one time in one shop for one guy!
Come on it's nowhere available :down:
11-04-2008, 06:05 PM
chew*

Just out of curiosity If deneb is 10-15% faster + if they were able to fix the errata wouldn't that equivalate to a 25% performance increase over they're current product?
Maybe i'm wrong, been out of the amd camp/loop for a while. I'm not sure how much the "patch" cost the phenoms in % / performance.
11-04-2008, 06:10 PM
v_rr

Quote:

Originally Posted by Epsilon84

I'm sure there will be wild debate as to where exactly Deneb falls on the IPC spectrum. I remember you doing the analysis on Agena a while back, but in the end there is no 'law' on what constitutes IPC and it really depends on the benchmark suite(s) used for comparison.

For example, in the recent 100 CPU roundup, Yorkfield is some ~30% faster than Phenom per clock: http://www.matbe.com/articles/lire/1...urs/page30.php

It´s precisely 25%.

It all depends on how AMD managed to tweak K10 in this shrink. The 3x bigger L3 cache will bring 6%-10% speed boost easily I think.
AMD talked albout IPC improvements in this K10.5. I expect improvements with all together from 5%-10%-15% depending on the benchmark.

In the end it will be faster then yorkfield? It will be very very difficult to happen.

There are 4 cenarios:
- K10.5 brings close to nothing performance improvement over K10 clock-for-clock. (not probable)
- K10.5 with same performance as Kentsfield (very probable)
- K10.5 sit between Yorkfield-Kentsfield. (very probable)
- K10.5 catch yorkfield (not probable)

So, lets just way for full reviews :)

Quote:

Originally Posted by AbelJemka

It was one time in one shop for one guy!
Come on it's nowhere available :down:

Forget him. In the time of 780G when where reports and people on foruns with some 780G he said that there was no 780G on market. Now there is one Corei7 in one forum and he says that it´s available.
11-04-2008, 06:48 PM
mAJORD

actually if you tally up results from Anandtech, Tech report, digit life, THG etc do any minor clock speed corrections required, you'll get

Kentsfield 9-11 % faster than agena
Yorkfield 14-17% faster than agena.

the matbe review is highly skewed towards Core 2 compared to 90% of the reviews out there.

I think it's because a couple of the benchmarks are very heavily core favoured.. i.e 50%+ / clock like Excel and it skews the results.. as does The Cool and Quiet issue that only Anandtech have discovered.. Which is of course AMD's own fault :down:

PS Just look at the Core 2 vs K8 results on Matbe, they're the same, don't match the accepted IPC difference over the last 2 yrs of benchmarks
11-04-2008, 06:49 PM
[XC] hipno650

i am thinking Deneb is going to end up a tiny bit faster than Kent. however what will make it an option will be prices and overclockablitiy. if they can have a $200 CPU that will hit from 3.5-4.0ghz on AIR then they will definetly be worth a look. However i just hope they don't have the same problem Phenom does that being how inconsistent they are in performance. some things are right there or ahead of Yorskfield clock per clock others they lose out to K8. time will tell i guess but they could be a real budget player again!
11-04-2008, 06:50 PM
Epsilon84

Quote:

Originally Posted by v_rr

It´s precisely 25%.

Yorkfield 3MB maybe? Compare the Q9450 to the 9950BE, its 32% faster, yes I realise its 2% higher in clockspeed too, which is why I said ~30%.

I agree with your possible scenarios btw, I think it'll be difficult for Deneb to exceed Yorkfield as Charlie claims but it could very well catch Kentsfield or slightly exceed it.
11-04-2008, 07:07 PM
Epsilon84

Quote:

Originally Posted by mAJORD

actually if you tally up results from Anandtech, Tech report, digit life, THG etc do any minor clock speed corrections required, you'll get

Kentsfield 9-11 % faster than agena
Yorkfield 14-17% faster than agena.

the matbe review is highly skewed towards Core 2 compared to 90% of the reviews out there.

I think it's because a couple of the benchmarks are very heavily core favoured.. i.e 50%+ / clock like Excel and it skews the results.. as does The Cool and Quiet issue that only Anandtech have discovered.. Which is of course AMD's own fault :down:

I agree to an extent, my point was to illustrate that there is no clear cut way to 'measure' IPC, it depends entirely on the mix of chosen benchmarks/apps. Therefore it is entirely possible for one site to have Deneb matching Yorkfield whilst another shows it lagging way behind, and we'll have the diehard fans from both camps touting the favourable reviews as gospel, isn't that how it always works? ;)

As for apps like Excel/Photoshop 'skewing' the results, maybe so, but where are the outlier benches where AMD doubles Core 2 performance outside of synthetic memory benches? I've never seen any, and I actually do think Excel and Photoshop has more relevance to most people than say Cinebench. Its a reviewers favourite and is obviously less 'skewed' towards Core 2 but in terms of usage I'd say Excel and Photoshop users outnumber Cinema4D users by orders of magnitudes... I guess reality is often skewed towards Core 2 as well. :shrug:
11-04-2008, 07:10 PM
JumpingJack

Quote:

Originally Posted by [XC] hipno650

i am thinking Deneb is going to end up a tiny bit faster than Kent. however what will make it an option will be prices and overclockablitiy. if they can have a $200 CPU that will hit from 3.5-4.0ghz on AIR then they will definetly be worth a look. However i just hope they don't have the same problem Phenom does that being how inconsistent they are in performance. some things are right there or ahead of Yorskfield clock per clock others they lose out to K8. time will tell i guess but they could be a real budget player again!

It's hard to tell ... the only leaked data set would suggest that Deneb is only clock for clock improvement over Agena of about 8% on average for desktop workloads, this is not enough to Yorksfield, but will put it about par with Kentsfield ... but power will be good and they will clock better.

AMD will compete well with price if they can bring it out in volume before Intel floods the market with lower priced Nehalem derivatives... so they should do very well, it will be a much more competitive part.

BTW: Rumor mill is grinding -- we may find out sooner than later: http://www.digitimes.com/news/a20081104PD203.html
11-04-2008, 08:00 PM
Macadamia

Gawd, I hate reading threads like these.

/goes off to watch moar breasts... err phones.
11-04-2008, 08:25 PM
Brother Esau

I will buy one for lower power consumption and reduced Heat out put.
11-05-2008, 01:24 AM
Shintai

Quote:

Originally Posted by chew*

Just out of curiosity If deneb is 10-15% faster + if they were able to fix the errata wouldn't that equivalate to a 25% performance increase over they're current product?
Maybe i'm wrong, been out of the amd camp/loop for a while. I'm not sure how much the "patch" cost the phenoms in % / performance.

The bug is already fixed in B3.
11-05-2008, 01:55 AM
AKM

Quote:

As for apps like Excel/Photoshop 'skewing' the results, maybe so, but where are the outlier benches where AMD doubles Core 2 performance outside of synthetic memory benches? I've never seen any, and I actually do think Excel and Photoshop has more relevance to most people than say Cinebench.

Data encryption

http://img142.imageshack.us/img142/2909/96484642el9.jpg
http://img142.imageshack.us/img142/9...jpg/1/w701.png

http://img359.imageshack.us/img359/4909/38350678xm1.jpg
http://img359.imageshack.us/img359/3...jpg/1/w741.png

http://www.guru3d.com/article/amd-ph...essor-tested/7

http://www.guru3d.com/article/intel-...-965-review/14
11-05-2008, 06:19 AM
demonkevy666

good reason for deneb 64 BIT

I want a 3.0ghz Deneb in 64 bit vista because I can't get 3.0ghz with out 1.625 volts, 60C at full load 41C idle.
11-05-2008, 07:00 AM
JumpingJack

Quote:

Originally Posted by chew*

Just out of curiosity If deneb is 10-15% faster + if they were able to fix the errata wouldn't that equivalate to a 25% performance increase over they're current product?
Maybe i'm wrong, been out of the amd camp/loop for a while. I'm not sure how much the "patch" cost the phenoms in % / performance.

This is why I never really like PR/Marketing from most all companies ... it seems they want to confuse as many people as possible.

AMD released Phenom with B2 stepping, the TLB errata, but when they demoed the chip to the press, the chip was not patched with the performance hit... what you saw was basically what you got.

The TLB patch turns off the tlb and the performance hit is massive, those numbers never made it into mainstream reviews, though some sites a few really good ones and some second tier, did a nice TLB vs non-TLB follow up.

B3 came out earlier this year, and fixed that TLB errata -- it was irrelevant since the B2 was basically rock solid, I have never seen the bug manifest itself (as is the case with most errata). B3 was equivalent IPC (Clock for clock) to B2 but with several bug fixes (bugs you would have never seen in the first place).
11-05-2008, 07:22 AM
G0ldBr1ck

I will be getting a couple of them regardless of any improvements just on the reason I have 2 unused 790fx boards. If 3.4-3.6ghz is possible on air it will be quite the upgrade from agena.
11-05-2008, 08:21 AM
Particle

Wasn't B3 more of a bandaid for the problem than an out-right fix? It was my understanding that the problem was eliminated, but the hardware didn't operate in the way that was originally intended. As in, B2 without the patch is slightly faster than B3.
11-05-2008, 08:34 AM
G0ldBr1ck

Quote:

Originally Posted by Particle

Wasn't B3 more of a bandaid for the problem than an out-right fix? It was my understanding that the problem was eliminated, but the hardware didn't operate in the way that was originally intended. As in, B2 without the patch is slightly faster than B3.

Thats news to me.
11-05-2008, 08:39 AM
Bellisimo

Quote:

Originally Posted by Particle

Wasn't B3 more of a bandaid for the problem than an out-right fix? It was my understanding that the problem was eliminated, but the hardware didn't operate in the way that was originally intended. As in, B2 without the patch is slightly faster than B3.

B3 is a faster rev and a complte fix for the TLB bug
11-05-2008, 08:49 AM
informal

Quote:

Originally Posted by Bellisimo

B3 is a faster rev and a complte fix for the TLB bug

Yep that's correct.
11-05-2008, 08:56 AM
Lightman

Quote:

Originally Posted by Bellisimo

B3 is a faster rev and a complte fix for the TLB bug

No it's not and Particle is more or less right.
B3 fix for TBL errata was a bypass. It didn't fix originally planned functionality. The fix brings minimal performance penalty, because TLB is flushed every time there can be some dirty data in it. I think Anand was doing more in-depth analysis.
11-05-2008, 09:02 AM
Particle

Can you refresh me on the details, Lightman? I can't remember why I thought what I said, just that that was my impression.

Show 100 post(s) from this thread on one page